I'm not sure from the title of the post what the question is. Never-the-less, I'll provide an answer. Up until now, there have problems when serialization code was spread accross and/or repeated in distinct runtime modules. Also, there was a problem when serialization was invoked simultaneously from different threads. Both these problems are related to the fact that the library has to keep a table of which types have the serialization code loaded. This has now been addressed in the recently checked in version in the trunk. Serialization code can be in dlls. Also code can be repeated accross DLLs. Personally, I would prefer to organize my code so that this doesn't happen but apparently some people found this inconvenient. I have tested this by hand with my VC 7.1 system. The tests consist of making separate DLLS and running the executable all from the same directory. I have been unable to figure out how to do such a thing with bjam so as far as linux is concerned its untested. Robert Ramey